Tone processing by non-tone language speakers

There have also been abundant studies that have tested the perception and production in beginning Mandarin L2 learners. For example, Wang et al. (1999, 2003) show that English learners of Mandarin improved their tone identification accuracy in mono- syllabic words from 69% to 90% after a two-week training. The training-induced im- provement also generalized to new words and speakers. Other than tones in isolated syllables, the perception of longer stimuli has also been tested. Hao (2012) found that both English and Cantonese learners of Mandarin performed better on monosyllabic tonal identification than on disyllabic identification. Both learner groups showed better performance on Mandarin tone mimicry than in tone identification and reading-aloud tasks. Mimicry only involved low-level auditory perception and articulation while the latter tasks required a more abstract representation of tones. This suggests that the main difficulty in tone learning is how to establish robust associations between pitch contours and tone categories.

More recently, the learning of new tonal categories by speakers without prior tone language learning experience has been tested in several phonetic training studies. Adopting different training paradigms (perception-only training, perception-plus-pro-

CHAPTER FIVE:TONAL AND SEGMENTAL INFORMATION IN WORD RECOGNITION 87

duction training and sound-to-word training), these studies examined the role of differ- ent modalities in training (Lu, Wayland, & Kaan, 2015), the distinction between reflect- ive and reflexive learning (Chandrasekaran, Yi, & Maddox, 2014), the contribution of individual variability in cue weighting in tone learning (Chandrasekaran, Sampath, & Wong, 2010), the effect of individual musical experience (Wong & Perrachione, 2007), as well as the influence of tonal context in tone learning (Chang & Bowles, 2015). Although the research questions varied across these studies, their results led to a convergent finding that naive non-native speakers of Mandarin can gain significant improvement in tonal identification and discrimination with a proper amount of training, and can learn to use tones in a lexically contrastive way. Some studies also found a training-induced change in participants’ neural system (Lu et al., 2015; Wong, Chandrasekaran, Garibaldi, & Wong, 2011). Since most of the studies mentioned above focused on the learning of lexical tones by naive non-native Mandarin speakers and beginning learners of Mandarin, the performance of advanced L2 learners and the developmental trajectory in the time course of tone acquisition have not been studied systematically. Moreover, L2 processing of tonal information has not been investigated using on-line methods. Therefore, the current study sets out to examine the role of tones and segments in auditory spoken word recognition using the Visual World Para- digm by monitoring the eye movements of both beginners and advanced Dutch learn- ers of Mandarin.

5.1.3 The present study

An eye-tracking experiment with Visual World Paradigm (VWP) was employed in the current study to test (1) the relative role of segmental and tonal information in lexical activation and selection by native speakers and Dutch learners of Mandarin, and (2) the developmental trajectory for Dutch learners of Mandarin in using segmental and tonal information effectively in spoken word recognition. Both beginners and advanced Dutch learners of Mandarin were recruited and native Mandarin speakers were tested as a control group.

VWP has typically been used in eye-tracking studies to investigate on-line auditory word recognition (Righi, Blumstein, Mertus, & Worden, 2009; Tanenhaus, Spivey-Knowlton, Eberhard, & Sedivy, 1995; also see a review in Heuttig, Rommers, & Meyer, 2011). Many related factors have been tested, such as the effect of frequency (Dahan, Magnuson, & Tanenhaus, 2001) and neighborhood density (Magnuson, Dixon, Tanenhaus, & Aslin, 2007). This paradigm has also been employed to test spoken word recognition by participants with language impairment (e.g., McMurray, Munson, & Tomblin, 2014; McMurray, Samelson, Lee, & Tomblin, 2010; Mirman, Yee, Blumstein, & Magnuson, 2011; Yee, Blumstein, & Sedivy, 2008). A recent study also demonstrated that this paradigm can provide reliable measurement for individual behavior (Farris- Trimble & McMurray, 2013).

In the VWP task, participants are presented with a display of four pictures and an auditory stimulus corresponding to one of these pictures, and they are asked to identify the word (i.e., the target) they heard with their eye movements being tracked during the whole process. In studies using VWP, the target word is always presented with a competitor which is phonologically similar to the target, and two phonologically unrelated distractors. For example, in Allopena, Magnuson, and Tanenhaus (1998),

TING ZOU:PRODUCTION AND PERCEPTION OF TONE BY DUTCH LEARNERS

88

participants were presented with a target word (e.g., beaker), a cohort competitor (e.g.,

beetle) (which shares the onset and vowel with the target), and two phonologically unrelated distractors (e.g., parrot, carriage). The results show that when hearing the instruction Pick up the beaker, participants tended to look at both beaker and beetle initially. As the word was unfolding, the target picture gained a greater proportion of looks. A similar effect was also found for the competitor that shared the rhyme with the target (e.g., speaker), indicating that listeners continuously extract segmental information from the acoustic signal and that phonologically similar lexical candidates can be gradually activated. This result supports the linking hypothesis between eye movements and lexical access, showing that fixations are time locked to the details of the speech input. This finding also closely matches the word recognition mechanisms posited by the TRACE model (McClelland & Elman, 1986), which assumes that when a word is heard, phonologically similar lexical candidates can be activated at any point in overlap with the speech input. Speech sounds presented at a feature layer can be mapped onto phoneme and word layer. With between-layer excitation and within-layer competitive inhibition, one word can be recognized among phonologically similar lexical candidates. The model suggests a lateral inhibition among lexical candidates, which can lead to a delayed activation of the correct target word. Studies using VWP have found evidence for this mechanism, in which the delayed activation is reflected as a delay in fixation on the target word (Dahan, Magnuson, Tanenhaus, & Hogan, 2001; Tanenhaus, Magnuson, Dahan, & Chambers, 2000).

The VWP has also been used in a recent study to examine the role of tonal information in spoken word recognition (Malins & Joanisse, 2010). In this study, Mandarin listeners were asked to identify the corresponding picture from four pictures while hearing a word. Both cohort competitor (sharing word initial phonemes and tone with the target) and segmental competitor (sharing segmental structure with the target and differing only in tone) caused slower eye movements to correct targets, indicating that tonal and segmental information play comparable roles in constraining lexical activation. Based on this finding and suggestions from previous studies (Malins & Joanisse, 2010; Ye & Connine, 1999; Zhao, Guo, Zhou, & Shu, 2011), tones have been incorporated into the TRACE model in a recent simulation of monosyllabic spoken word recognition of Mandarin Chinese (Shuai & Malins, 2017).

The processing of segmental and tonal information in word recognition by Dutch learners of Mandarin is tested in the current study. To test the competition effect of segmental versus tonal cues, the target words are presented with different types of competitors: cohort competitors sharing the initial consonant and tone with the target (e.g., che1 ‘car’; for the target of chuang1 ‘window’); rhyme competitors with rhyme and tone overlap (e.g., guang1 ‘light’); segmental competitors with complete segmental overlap (e.g., chuang2 ‘bed’), and tone competitors with tone overlap (e.g., ji1

‘chicken’). The probability of fixation to the target and competitors was recorded since it may reflect the activation of the corresponding items.

CHAPTER FIVE:TONAL AND SEGMENTAL INFORMATION IN WORD RECOGNITION 89 5.2 Method